Sourcing the Arab Spring: A Case Study of Andy Carvin’s...

Journal of Computer-Mediated Communication

Sourcing the Arab Spring: A Case Study of AndyCarvin’s Sources on Twitter During the Tunisianand Egyptian Revolutions∗

Alfred Hermida

Graduate School of Journalism, University of British Columbia, 6388 Crescent Road, Vancouver, B.C. V6T 1Z2,Canada

Seth C. Lewis

School of Journalism & Mass Communication, University of Minnesota–Twin Cities, 111 Murphy Hall,206 Church Street SE, Minneapolis, MN 55455, USA

Rodrigo Zamith

School of Journalism & Mass Communication, University of Minnesota–Twin Cities, 111 Murphy Hall,206 Church Street SE, Minneapolis, MN 55455, USA

News sourcing practices are critical as they shape from whom journalists get their information andwhat information they obtain, mostly from elite sources. This study evaluates whether social mediaplatforms expand the range of actors involved in the news through a quantitative content analysisof the sources cited by NPR’s Andy Carvin on Twitter during the Arab Spring. Results show that, onbalance, nonelite sources had a greater representation in the content than elite sources. Alternativeactors accounted for nearly half of the messages. The study points to the innovative forms of produc-tion that can emerge with new communication technologies, with the journalist as a central nodetrusted to authenticate and interpret news flows on social awareness streams.

Key words: Andy Carvin, Arab Spring, gatekeeping, journalism, sourcing, Twitter.

doi:10.1111/jcc4.12074

This paper presents a case study on the use of sources by National Public Radio’s Andy Carvin on Twitterduring key periods of the 2011 Tunisian and Egyptian uprisings. Carvin, a social media strategist at NPRin the US, emerged as a key broker of information on Twitter during the Arab Spring.1 Through Twitter,Carvin would often link to images from demonstrators, curate a range of discussion and opinion aboutevents, and frequently ask his followers (then about 50,000 strong) to help him make sense of the bits ofinformation he encountered. This study examines the different actor types on the social media platformto reveal patterns of sourcing of information used by Carvin in order to further an understanding ofhow sourcing is evolving in an era of networked digital media—a space that, by its nature, allows fornew research possibilities in tracking the influence of sources.

∗Accepted by previous editor Maria Bakardjieva

Journal of Computer-Mediated Communication 19 (2014) 479–499 © 2014 International Communication Association 479

The interplay between journalists and sources is a significant factor in affecting what and whomakes the news. Sources help to shape how events and issues are reported, influencing the public’sunderstanding of the world. Studies on journalistic practices have highlighted key challenges in newsroutines, including a limited diversity of news sources and a reliance on those with institutional power,such as government officials, police officers or business leaders. The open nature of social mediatechnologies could, in theory, foster greater pluralism in media discourse by providing channels for agreater number and diversity of news sources.

The availability of the messages sent by Andy Carvin on Twitter offers an opportunity to analyzehis choice of sources on the Arab Spring, investigate the dynamics of discourse across institutional andalternative sources, and chart the predominance of voices on Twitter. This case study, built on a quan-titative content analysis of his most frequently cited sources, provides insight into the process throughwhich Carvin, as a member of the professional news media, negotiates information gathering and dis-semination in this emerging social environment.

The Gatekeeping of News Sources

News sources are a critical element in the practice of journalism as it shapes from whom journalistsget their information and what type of information they obtain (Carlson, 2009; Gans, 1979). Sourcinginvolves making decisions on who is included or excluded as an actor in the media. The sources citedin the media do more than denote events and issues. They ascribe meaning to events, shaping publicperception and understanding. Studies into the production of news have shown that journalists seekto cite sources considered authoritative as they hold certain credentials in society (Ericson, Baranek,& Chan, 1989; Gans, 1979; Tuchman, 1978). These credentials stem from bodies holding positions ofpower, such as government or police, or representing significant segments of society, such as in business.Numerous studies have identified how government officials dominate sources (Brown, Bybee, Wearden,& Straughan, 1987; Sigal, 1973). Hall, Critcher, Jefferson, Clarke, & Roberts (1978) argue that elite sourcesare at the top of a hierarchy of credibility, and as a result, are primary definers who shape the news agendaand interpretative approach used by journalists. For Hall et al., the deadlines of news production and theprofessional commitment to impartiality and objectivity “combine to produce a systematically structuredover-accessing to the media of those in powerful and privileged institutional positions” (1978, p. 58). Notonly do nonelite sources make up a small minority of news sources, Ericson et al. argue they are used toinspire “fear and loathing” (1989, p. 1), reinforcing the authority of elites.

The use of elite sources by journalists further enhances the credibility of these sources. As Tuchman(1978, p. 210) notes, “by identifying centralized sources of information as legitimated social institutions,news organizations and newsworkers wed themselves to specific beats and bureaus. Those sites are thenobjectified as the appropriate sites at which information should be gathered.” The hierarchy of sourcesis replicated in alternative media, where a counterelite of source types “preserves the dominant modelof sourcing in its assumptions about power, legitimacy and authoritativeness” (Atton & Wickenden,2005, p. 357).

News practices reinforce what Becker defines as a hierarchy of credibility, where elites are presumedto have greater authority in defining “the way things really are” (1967, p. 241). Reporters develop routinesto manage organizational limits and maximize efficiency (Gans, 1979; Tuchman, 1978). Journalists areexpected to have a range of sources, yet come across operational impediments, including the geographicand social proximity of the source. Recent developments, such as the acceleration of the news cycle andthe corresponding shortening of the publication cycle, have affected sourcing, with greater reliance onsecondhand sources, such as news agency content or already published news stories (Boczkowski, 2010;

480 Journal of Computer-Mediated Communication 19 (2014) 479–499 © 2014 International Communication Association

Thurman & Myllylahti, 2009). Shrewd sources understand the limitations facing journalists and employtactics to satisfy a reporter’s need for information, hence increasing their chances of being cited. Gans(1979) describes the process as a tug-of-war between sources seeking to shape the content and directionof news, and journalists striving to obtain the details needed.

The credibility, or perceived credibility, of a source is a key determinant in the tug-of-war (Gans,1979; Reich, 2011). While source type and affiliation affect how journalists assign credibility, operationallimitations mean reporters develop ongoing relationships with sources. The more familiar journalists arewith a source, the more likely they are to be considered credible. Engagement influences credibility andthe likelihood of being used as a source. As Gans notes, “those they talk with frequently can be evaluatedover time, which is another reason why story selectors prefer regular sources” (1979, pp. 129–130).In his study on sourcing and credibility, Reich (2011) found that journalists tended to depend almostexclusively on a core set of sources who had proved their trustworthiness in the past.

The selection of news sources, therefore, is an essential form of gatekeeping—the process throughwhich journalists filter vast quantities of information to distill a narrow set of news reports for a givenday (Shoemaker, 1991). The gatekeeping metaphor (White, 1950) has formed the theoretical basis for awide range of mass communication scholarship, including many contemporary accounts of journalists’efforts to “guard open gates” (Singer et al., 2011) as news organizations increasingly engage participa-tory forms of news (Lewis, Kaufhold, & Lasorsa, 2010). In their seminal meta-analysis, Shoemaker andVos (2009) propose that scholars examine the influences shaping gatekeeping at five levels of analysis:individual (e.g., personal background), routines (e.g., work patterns), organizational (e.g., media own-ership), social institutional (e.g., extraorganizational forces such as public relations), and social system(e.g., ideology) (cf., Shoemaker & Reese, 1996). Moreover, Shoemaker and Vos suggest that a revisedversion of the gatekeeping model should give primacy not only to journalists (media channel) and theirinformation (source channel) but should acknowledge the growing impact of user-driven information(audience channel) in shaping the media and source channels (see Figure 9.1 on p. 125). Additionally,Shoemaker and Vos contend that research should better account for the agency of individual gatekeep-ers, given that “the sociological turn in gatekeeping studies has left Mr. Gates as a minor character in theselection of news” (p. 134). This study attempts to address this concern and build upon their frameworkby (1) examining an individual gatekeeper whose agency is manifest in his publishing to Twitter withoutoversight from senior editors; (2) simultaneously addressing the social institutional level of influencefrom sources; and (3) exploring the extent to which journalists operating on social media may performtheir gatekeeping function differently in relation to traditional or nontraditional types of sources.

Journalistic Sourcing and the Social WebWeb 2.0 technologies, often referred to as social media, offer broad opportunities for individuals to par-ticipate in the observation, filtering, distribution and interpretation of news. The negotiation betweenjournalism and social media as structuring and/or shaping technology is a key point for understandingits role in influencing established norms, practices and routines. Social media allow for new relations thatpotentially disrupt hierarchical structures and erode the traditional distinction between the producerand consumer of news and information.

Services like Twitter facilitate the instant, digital dissemination and reception of short fragments ofdata from sources both inside and outside the framework of established journalism. The free service hasgrown as a network for real-time news and information since its creation in 2006, shaping how newsis gathered, distributed, and received (Bruno, 2011; Hermida, 2010; Lasorsa, Lewis, & Holton, 2012;Newman, 2009). In its short lifespan, Twitter has attracted attention for its role in the reporting of majorevents, such as the terrorist attacks in Mumbai in November 2008, the protests following the Iranian


election in June 2009, the earthquake in Haiti in 2010, and uprisings in Middle East (Bruno, 2011; Kwak,Lee, Park, & Moon, 2010).

Twitter describes itself as “a real-time information network that connects you to the latest informa-tion about what you find interesting” (Twitter, n.d.). By March 2012, it reported 140 million active usersand 340 daily million messages (Twitter, 2012). Hermida (2010) has described the flows of news andinformation on Twitter as ambient journalism. Ambient journalism frames Twitter as a social awarenesssystem that delivers a fragmented mix of information, enlightenment, entertainment, and engagementfrom a range of sources. In certain types of situations, Twitter users take on the role of social sensorsof the news (Sakaki, Okazaki, & Matsuo, 2010), with the network serving a channel for breaking newsalerts and subsequently for a stream of real-time data as events unfold.

As a result, Twitter has been promptly adopted in newsrooms as a mechanism for user-generatedcontent, often filling the news vacuum that can follow the immediate aftermath of a breaking news eventby sourcing eyewitness accounts, photos, and video from social media. This has given rise to the role ofthe journalist as curator who filters, selects and contextualizes copious amounts of real-time informa-tion on the fly (Bruno, 2011; Newman, 2009). The role of the journalist is reframed as a professional who“lays bare the manner through which a news story is constructed, as fragments of information are con-tested, denied or verified” (Hermida, 2012, p. 8). The technical architecture of Twitter presents distinctresearch opportunities to study the relationship between the journalist and sources, offering insightsinto the engagement with sources and the subsequent broadcast of information from these sources. Theinteractions between a journalist and a source are traditionally hidden from public view, making it dif-ficult to assess whom a journalist has engaged with in the process of collecting information. Researchon sourcing focuses on who is quoted by journalists as this can be measured by counting the number ofcitations in a news article or broadcast. However, the interactions between a journalist and a source arecaptured by the @mentions mechanisms on Twitter, revealing how a reporter engages with sources togain information, background, and context. The sources cited are captured by the retweet mechanism,when a journalist broadcasts a message from a source.

Networked and distributed social media platforms potentially expand the range of actors involvedin the construction of the news. Yet studies indicate that the ability of media audiences to participatein the processes of news production within professional publications has been severely circumscribed(Lasorsa et al., 2012; Lewis, 2012; Singer et al., 2011). Bruno’s study of the coverage of three major newsoutlets of the 2010 Haiti earthquake suggests an opportunistic model at play, rather than a desire torepresent a broad spectrum of voices. Bruno found a significant reliance of social media content by theBBC, The Guardian, and CNN in the first 24 hours of the natural disaster. But the use of social mediacontent fell dramatically once the BBC and CNN had their own teams in Haiti. Bruno concluded thatonly “The Guardian seems to have embraced an editorial policy more open and consistent with regardto the diversity of online voices” (2011, p. 63).

Social media are attractive to activists as they can offer alternative platforms for public commu-nication that bypass the gatekeeping of traditional media (Bruns, 2008). In the coverage of protestmovements, the sourcing practices of the mainstream media shape the nature and tone of coverage.Journalists rely on institutional actors perceived as authoritative sources, such as police and officials,marginalizing alternative voices that are seen as deviant (Bennett, 1988; Hall et al., 1978). In their studyon social media during the G20 protests in Toronto in 2010, Poell and Borra (2011) suggested thatTwitter held the most promise for crowdsourced alternative reporting. However, they noted that thereporting was led by a small number of users who had emerged as an elite set of voices through thepractice of retweeting. Moreover, Poell and Borra noted that the narrative on social media mirroredmainstream reporting on the violence during the protests as though activists focused on reports ofviolence by the police, rather than protesters.


Emerging research suggests that social media, and more specifically Twitter, provides a platform forthe coconstruction of news by journalists and activists. In a study of tweets during the Tunisian andEgyptian uprisings, Lotan et al. (2011) found that both journalists and activists were key informationsources. Activists were the top type of source cited on Twitter for Tunisia, whereas journalists becamethe main type of source for Egypt. The findings suggest that activists filled a news vacuum in Tunisia,a traditionally unreported country, as the protests unfolded and the international media started to playmore attention. Egypt, in contrast, has tended to be better covered in the mainstream media in the past,and the protests garnered greater attention as they followed the ones in Tunisia.

Other studies of the protests that reshaped the Middle East in 2011 have highlighted how socialmedia can give voice to a set of alternative sources. In their study of tweets using the #Egypt hash-tag, Papacharissi and de Fatima Oliveira (2012) found that the more prominent voices on the Egyptianuprising belonged to elite news organizations and specific individuals. They found that, together withmainstream media journalists, there was a parallel and significant set of voices consisting of bloggers,activists and intellectuals involved in advocacy. This alternative set of elite voices was crowdsourcedthrough the mechanisms of social media “that reward those more involved in mobilization, and thereporting and curating of information, online and offline,” (2012, p. 14). Papacharissi and de FatimaOliveira suggest that the stream of news on Twitter combined news, opinion and emotion, pointing to amix of old and newer news values.

The Case of Andy CarvinTo examine journalistic sourcing dynamics in a fluid news space like Twitter, and in particular howthese dynamics played out in the Arab Spring, we have chosen to study Carvin’s work at the height ofthe Tunisian and Egyptian uprisings. His coverage on Twitter—spanning upwards of 16 hours a day, 7days a week—featured hundreds of tweets per day (Farhi, 2011). As Carvin emerged as a central node inthe information network on the Arab Spring, his peers took notice. Summing up the reaction in mediacoverage after Carvin was featured in The New York Times, the Guardian, and The Washington Post,the Columbia Journalism Review called Carvin’s Twitter feed a “living, breathing real-time verificationsystem” and a “must-read newswire” (Silverman, 2011, para. 1). His prominence on Twitter resultedin a 2012 Shorty Award, which recognizes the best producers of short-form content on social media(PRWeb, 2012).

The literature on journalism and sourcing, the emergence of social media, and the particular caseof Andy Carvin raise significant questions. First, given how the choice of sources influence how eventsare reported, our research seeks to understand how certain actors gain more attention than others inthe process of gathering and filtering news. Second, the transparency of digital networks, and Twitterin particular, facilitate the work of identifying how such sourcing occurs. Third, the unique context ofCarvin’s role may point to a new kind of journalistic style emerging in social spaces, where reporters relyon a potentially broader array of sources, from citizens to individual activists to institutional bodies. Thefirst two items are addressed in the research questions that follow; the third is the focus of the Discussionsection that follows our findings.

Research QuestionsOur primary concern is the nature of Carvin’s sources: the type of actors involved, and the prominencethat certain actors achieve relative to others. This leads us to ask:

RQ1. What types of sources are most prominent in Carvin’s coverage of the Egyptian and Tunisianrevolutions?


Secondly, because the networked and public architecture of Twitter presents an opportunityto observe the online interactions between journalists and sources—e.g., @mentions as a form ofengagement, and retweets as a form of broadcasting—we seek to investigate the nature of Carvin’ssourcing activity, and assess how such activity is associated with source type prominence.

RQ2. How does the relative prominence of source types vary according to the sourcing practicesthat Carvin employed during these periods?

Methods

SampleThe data for this study came from a dataset, provided by Carvin and obtained from Twitter, that includedall of Carvin’s tweets—more than 60,000 of them—posted between December 1, 2010 and September16, 2011. The researchers developed a computer program to parse this data and systematically categorizethem based on several criteria (for details, see Lewis, Zamith, & Hermida, 2013).2 All of the tweetsappearing between January 12 and January 19, and from January 24 to February 13, were subsequentlyisolated by the researchers. The first period covers the major portion of Tunisian demonstrations leadingto the fall of President Ben Ali, and the second covers the Egyptian protests and subsequent resignationof President Hosni Mubarak. These choices of dates were made according to an analysis of news timelinesfor major events during the Arab Spring,3 and to correspond with similar time frames used in previousresearch of this kind (e.g., Lotan et al., 2011). Carvin tweeted a total of 411 times during the Tunisian timeperiod, including in his tweets 191 unique sources. During the Egyptian time period, Carvin tweeted5,290 times and included 1,156 unique sources in his tweets.

To create a comparable and sufficiently large, yet manageable, sample, the researchers opted tocode all profiles that accounted for 0.09% or more of the retweeted sources or 0.25% or more of thenon-retweeted sources.4 This yielded 330 unique sources, with 190 sources appearing in the Egyptsample, 172 in the Tunisia sample, and 32 sources overlapping both samples. The Twitter profiles forthese sources were then systematically downloaded by the researchers on December 6, 2011. A total ofeight profiles could not be obtained since they had been either deleted or protected from public view,resulting in a final sample of 322 sources: 185 for Egypt and 168 for Tunisia, with 31 sources overlapping.

Coding Instrument

This study was primarily concerned with two key variables: (a) the type of interaction that occurred ina tweet and (b) the type of source being interacted with. To determine the type of interaction occurringin each of Carvin’s tweets, a computer program was used to systematically ascertain whether tweetswere broadcasting information or engaging sources. Tweets deemed to be broadcasting information werethose that included a retweet, and were identified through the presence of the text “RT @” in the tweet,with the source being retweeted determined as the handle immediately adjacent to the expression. Incases where the text “RT @” appeared multiple times in one tweet, the first instance was given priority.Tweets that did not include the text “RT @” but did include a source (“@”) were deemed to representengagement, with all sources appearing in the tweet listed as having been engaged by Carvin. Thus,retweets were classified as broadcasting and non-retweets as engaging.

The source type variable was initially assessed by two independent coders, who were blind to theresearch questions, and later re-evaluated by the researchers. All sources were coded using a customizedelectronic coding interface developed by one of the researchers. In this system, the source’s Twitterprofile page appeared adjacent to the electronic code sheet. This eliminated data entry error by


automatically transferring entries into a relational database, removing the need for human intervention.It also helped reduce coder error by having options and categories presented as labels, with the systemautomatically converting selections into numerical values after each submission. Coders were instructedto rely primarily on the data from the source’s stored profile, although they were also permitted to accessexternal sources like the source’s current Twitter profile, associated blogs and personal websites, andLinkedIn profile.

The source type classifications were adapted from Lotan et al. (2011). They included: affiliatedactivists, nonaffiliated activist, bloggers, bots, celebrities, digerati, mainstream media employees, main-stream media organizations, mainstream new media organizations, nonmedia activist organizations,nonmedia nonactivist organizations, political actors, researchers, and any other type of account. For adescription of each source type, see Table 1.

Intercoder ReliabilityTo assess intercoder reliability, the independent coders double-coded 46 randomly selected sources(14.3% of the sample) that appeared within the dataset but outside of the sample. To determine reli-ability, the researchers used Scott’s Pi, which corrects for chance agreement (Scott, 1955). The coefficientfor the source type variable was .72, thereby exceeding the minimum bound of .70 suggested by Riffe,Lacy, and Fico (2005).

In light of the complexity of the source type variable—indeed, as in the Lotan et al. (2011) study,several sources were found to span several categorizations, presenting significant challenges—all profileswere subsequently reviewed by the researchers to ensure validity. When a researcher disagreed with acoder’s classification—which occurred almost exclusively in the more ambiguous cases—that profilewas reviewed by all three researchers simultaneously and recoded through consensus-building. A totalof 60 sources (18.6%) were recoded in this manner.

Results

Source Type ProminenceThe relative prominence of source types (RQ1) was determined first by examining source representationin the sample (see Table 2). For instance, in the overall population of individual sources, mainstreammedia employees accounted for the largest group by far (26.7%). They were likewise the largest group byproportional representation within the Egypt (33.0%) and Tunisia (20.2%) subsamples. A closer look atproportional differences between countries reveals subtle changes as Carvin moved from the Tunisianrevolution to the Egyptian revolution. For example, there was a substantial uptick in the representation ofmainstream media employees and nonaffiliated activists, and a decline in the representation of digerati.

Proportional representation in the sample, however, is just one form of prominence; a more impor-tant measure is the relative frequency of tweet mentions given to source types (see Table 2). On thisscore, a different picture emerges: Compared to other individual source categories, nonaffiliated activistsaccounted for the greatest single share of tweet mentions, overall (35.3%) and for Egypt (37.5%). Mean-while, mainstream media employees had the second-largest proportion for Egypt (30.1%) and the mostfor Tunisia (23.2%). Notably, digerati had a 16.3% share for Tunisia, but that fell to 3.8% in the Egyptsample.

A third step in this analysis of source type prominence was to organize the discrete source typecategories originally deployed (see Table 1) into a more coherent, three-part scheme: Mainstream Media,Institutional Elites, and Alternative Voices (see Table 2). This arrangement allowed for a more focusedcomparison of the extent to which Carvin drew upon traditional sources for news (fellow journalists,


Table 1 Source Type Definitions

Source Type Description Example

Activist (Affiliated) Individuals who either self-identify as anactivist or who appear to be tweeting purelyabout activist topics, and affiliate themselveswith an advocacy group or organization.

@RachelPerrone

Activist (Non-Affiliated) Individuals who either self-identify as anactivist or who appear to be tweeting purelyabout activist topics, but do not affiliatethemselves with an advocacy group ororganization.

@Elazul

Blogger Individuals who post regularly to anestablished blog, and who appear to identifyas a blogger on Twitter.

@paulseaman

Bots Accounts that appear to be an automatedservice tweeting consistent content, usuallyin extraordinary volumes.

@toptweets

Celebrity Individuals who are famous for reasonsunrelated to technology, politics, or activism.

@Alyssa_Milano

Digerati Individuals who have worldwide influence insocial media circles and are, thus, widelyfollowed on Twitter.

@ev

Mainstream MediaEmployees

Individuals employed by MSM organizations,or who regularly work as freelancers forMSM organizations.

@camanpour

Mainstream MediaOrganization

News and media organizations that have bothdigital and non-digital outlets.

@NYTimes

Mainstream New MediaOrganization

Blogs, news portals, or journalistic entities thatexist solely online.

@visionOntv

Non-Media Organization(Activist)

Groups, companies, or organizations that arenot primarily news-oriented and openlyadvocate a point of view or support a cause.

@amnesty

Non-Media Organization(Non-Activist)

Groups, companies, or organizations that arenot primarily news-oriented and do notopenly advocate a point of view or support acause.

@instagram

Political Actor Individuals who are known primarily for theirrelationship to government.

@jeanmarcayrault

Researcher An individual who is affiliated with a universityor think-tank and whose expertise seems tobe focused on Middle East issues.

@jeffjarvis

Other Accounts that do not clearly fit into anycategory.

@AngelinesMoncha

Note: Source type categories adapted from Lotan et al. (2011).


Table 2 Relative Prominence of Source Types—by Individual Source Type and Recoded Group SourceType—in Andy Carvin’s Tweets During the Egyptian and Tunisian Uprisings in 2011

Proportion of mentions in Carvintweets in the sample:

Source Type(organized bygroup categorization)

Overallproportion(N= 322)

Egyptproportion(N= 185)

Tunisiaproportion(N= 168)

Overall(N= 3,623)

Egypt(N= 3,291)

Tunisia(N= 332)

Mainstream MediaEmployees

26.7% 33.0% 20.2% 29.4% 30.1% 23.2%

Mainstream Media Org. 4.3% 4.3% 4.8% 2.0% 1.8% 3.6%Mainstream New Media

Org.1.2% 1.6% .6% 1.2% 1.3% .3%

Mainstream Media(group subtotal)

32.3% 38.9% 25.6% 32.6% 33.2% 27.1%

Digerati 12.4% 9.7% 17.9% 4.9% 3.8% 16.3%Researchers 6.8% 7.0% 7.1% 5.6% 5.3% 7.8%Non-Media Org

(Non-Activist)2.5% .5% 4.8% .5% .1% 4.2%

Celebrities 1.2% 1.6% 1.2% .5% .5% .9%Political Actors .9% .5% 1.2% .2% .1% .6%Institutional Elites (group

subtotal)23.9% 19.3% 32.1% 11.7% 9.8% 29.8%

Non-Affiliated Activist 14.6% 18.4% 10.1% 35.3% 37.5% 13.6%Bloggers 6.5% 4.3% 7.7% 7.6% 7.5% 9.0%Affiliated Activist 3.4% 2.7% 4.2% 3.9% 3.5% 7.8%Non-Media Org

(Activist)1.9% 2.2% 1.2% 1.5% 1.6% .6%

Alternative Voices (groupsubtotal)

26.4% 27.6% 23.2% 48.3% 50.1% 31.0%

Other 17.4% 14.1% 19.0% 7.3% 6.8% 12.0%Sample (Total) 100.0% 100% 100% 100% 100% 100%

Note: Totals may not add to 100% because of rounding error. The Ns for Egypt proportion and Tunisiaproportion exceed the Overall proportion because of source overlap.

researchers, institutions, and other elite actors) and nontraditional sources (activists and bloggers). Thisgroup-level comparison makes it clear that Alternative Voices enjoyed an outsized influence in Carvin’scoverage; while they accounted for barely a quarter of his sources overall, they nonetheless receivedroughly half of all tweet mentions in the sample—far more than either Mainstream Media (32.6%) orInstitutional Elites (11.7%).

Sourcing Practices and ProminenceThe second research question sought to examine the nature of Carvin’s sourcing practices—whether inthe form of broadcasting (RTs) or engaging (non-RTs)—and assess how such activities are associatedwith source type prominence. The data were broken down by revolutionary period and by sourcingpractice within each period (i.e., Egypt RTs, Egypt non-RTs, Tunisia RTs, Tunisia non-RTs). Sources in


Table 3 Top-25 Retweeted Sources, Egypt

# Handle Name Source Category # of Tweets

1 @sultanalqassemi Sultan Al Qassemi Alternative Voices 1612 @bencnn Ben Wedeman Mainstream Media 993 @monasosh Mona Seif Alternative Voices 924 @mosaaberizing Mosa’ab Elshamy Alternative Voices 875 @evanchill Evan Hill Mainstream Media 806 @dima_khatib Dima Khatib Alternative Voices 727 @riverdryfilm Omar Robert Hamilton Alternative Voices 668 @3arabawy Hossam el-Hamalawy Alternative Voices 659 @weddady Nasser Weddady Alternative Voices 6410 @alaa Alaa Abd El Fattah Alternative Voices 6311 @ramyyaacoub Ramy Yaacoub Institutional Elites 5612 @zeinobia “Zeinobia” Alternative Voices 5613 @ghonim Wael Ghonim Alternative Voices 5214 @nickkristof Nicholas Kristof Mainstream Media 5215 @sharifkouddous Sharif Kouddous Alternative Voices 5116 @laraabcnews Lara Setrakian Mainstream Media 4917 @monaeltahawy Mona Eltahawy Mainstream Media 4818 @gsquare86 Gigi Ibrahim Alternative Voices 4719 @egyptocracy “Egyptocracy” Alternative Voices 4320 @nadiae Nadia El-Awady Alternative Voices 4221 @waelabbas Wael Abbas Alternative Voices 3822 @bloggerseif Ali Seif Alternative Voices 3723 @jan25voices “Jan25 Voices” Alternative Voices 3624 @sandmonkey Mahmoud Salem Alternative Voices 3525 @sherinet Sherine Tadros Mainstream Media 35

Note: The Mainstream Media category comprises of mainstream media organizations, mainstream newmedia organizations, and mainstream media employees. The Alternative Voices category comprises ofbloggers, non-media activist organizations, affiliated activists, and non-affiliated activists. The Institu-tional Elites category comprises of non-media non-activist organizations, political actors, celebrities,researchers, and digerati. The Other category comprises of bots and all other source types.

this context were considered at the level of the individual (single Twitter user) and the recoded three-partgroup structure noted above.

For prominence among individual sources, Tables 3–6 display a list of the top 25 sources for eachof the four periods and sourcing practice classifications. Overall, there was little overlap across thelists,5 suggesting that, at least among go-to individuals, Carvin turned to a different set of sourcesfor each revolutionary period—and even, it would appear, different leading sources.6 These lists alsoserve to illustrate the dramatic increase in Carvin’s use of Twitter between the Tunisian and Egyptianperiods, judging by the number of mentions required to crack the top 25 groups for each country.These tables also reveal the outsized influence of certain individuals, such as Sultan Al Qassemi (@sul-tanalqassemi), whose relative prominence among Egypt retweeted sources was unrivaled across theentire sample.


Table 4 Top-25 Non-Retweeted Sources, Egypt

# Handle Name Source Category # of Mentions

1 @ghonim Wael Ghonim Alternative Voices 262 @alaa Alaa Abd El Fattah Alternative Voices 193 @sultanalqassemi Sultan Al Qassemi Alternative Voices 194 @pubmedia “#PubMedia Chat” Mainstream Media 185 @shanestolar Shane Stolar Other 186 @wjchat “wjchat” Mainstream Media 187 @evanchill Evan Hill Mainstream Media 178 @mathewi Mathew Ingram Mainstream Media 159 @webjournalist Robert Hernandez Institutional Elites 1310 @weddady Nasser Weddady Alternative Voices 1311 @mosaaberizing Mosa’ab Elshamy Alternative Voices 1212 @3arabawy Hossam el-Hamalawy Alternative Voices 1113 @dima_khatib Dima Khatib Alternative Voices 1114 @monasosh Mona Seif Alternative Voices 1015 @yayayarndiva P. Mimi Poinsett Other 1016 @manal Manal Hassan Alternative Voices 917 @sandmonkey Mahmoud Salem Alternative Voices 918 @monaeltahawy Mona Eltahawy Mainstream Media 819 @ramyyaacoub Ramy Yaacoub Institutional Elites 820 @ajenglish Al Jazeera English Mainstream Media 721 @antderosa Anthony De Rosa Mainstream Media 722 @ivancnn Ivan Watson Mainstream Media 723 @jan25voices “Jan25 Voices” Alternative Voices 724 @jeffjarvis Jeff Jarvis Institutional Elites 725 @jilliancyork Jillian C. York Alternative Voices 7

Note: The Mainstream Media category comprises of mainstream media organizations, mainstreamnew media organizations, and mainstream media employees. The Alternative Voices category com-prises of bloggers, non-media activist organizations, affiliated activists, and non-affiliated activists.The Institutional Elites category comprises of non-media non-activist organizations, political actors,celebrities, researchers, and digerati. The Other category comprises of bots and all other sourcetypes.

Furthermore, by comparing relative proportions among source categories in these tables, additionaldistinctions come into focus. Whereas the spread of actor types is more mixed in the case of Tunisia,Alternative Voices dominate the top 25 for Egypt, accounting for 18 on the list of RT sources and 12 onthe list non-RT sources. On the flip side, Institutional Elites were top among Tunisia non-RT sources,with 9 spots, but accounted for only 3 of the top 25 non-RT sources in Egypt—and only 1 of the top 25for Egypt RT.

To gauge the prominence of source type groups across the whole sample, descriptive statistics wereanalyzed (see Table 7). These reinforce the relative outsized impact of Alternative Voices in Egypt. Thisis particularly true in the dramatic differences across median number of tweet mentions for differentgroups on the Egypt RT dimension, as revealed in a median test: 𝜒2 (2, N = 141) = 20.12, p < .001,


Table 5 Top-25 Retweeted Sources, Tunisia

# Handle Name Source Category # of Tweets

1 @dima_khatib Dima Khatib Alternative Voices 62 @jrug Jonathan Rugman Mainstream Media 63 @nawaat Nawaat de Tunisie Alternative Voices 44 @sultanalqassemi Sultan Al Qassemi Alternative Voices 45 @achrisafis Angelique Chrisafis Mainstream Media 36 @brian_whit Brian Whitaker Mainstream Media 37 @ibnkafka “ibnkafka” Alternative Voices 38 @ifikra Sami Ben Gharbia Alternative Voices 39 @lukebozier Luke Bozier Institutional Elites 310 @shadihamid Shadi Hamid Institutional Elites 311 @alanfisher Alan Fisher Mainstream Media 212 @bengacem Leila Ben-Gacem Other 213 @edwebb Ed Webb Institutional Elites 214 @harhour_l “Harhour” Other 215 @jeffjarvis Jeff Jarvis Institutional Elites 216 @jilliancyork Jillian C. York Alternative Voices 217 @mmm Moh’d M. Meddah Alternative Voices 218 @monaeltahawy Mona Eltahawy Mainstream Media 219 @niemanlab Nieman Lab Institutional Elites 220 @nprnews “NPR News” Mainstream Media 221 @sgardinier Suzanne Gardinier Alternative Voices 222 @simsimt Usamah M. Other 223 @techsoc Zeynep Tufekci Institutional Elites 224 @weddady Nasser Weddady Alternative Voices 225 @_niss Aniss Bouraba Other 1


Cramer’s V = .38. The picture for Tunisia is more muddled, primarily because of the small number oftweets posted during the period (see the Sum column).

In the absence of normal distributions,7 a Kruskal-Wallis test was conducted to evaluate differencesamong the three group types (Mainstream Media, Institutional Elites, and Alternative Voices) on medianchange in the dependent variable (the number of tweet mentions across each of the four sourcing activityclusters). Results are displayed in Table 7. The test was significant for Egypt RT and non-RT. For theEgypt retweet cluster, the effect size, or the proportion of the variability in the ranked dependent variableaccounted for by source type independent variable, was .12, indicating a moderate relationship betweensource type and change in the number of tweet mentions. The effect size for Egypt non-retweet wasweaker, but still noteworthy, at .06.

Follow-up tests were conducted to assess pairwise differences among the three groups in each ofthe two Egypt clusters, controlling for Type I error across the tests by incorporating the Bonferroni


Table 6 Top-25 Non-Retweeted Sources, Tunisia

# Handle Name Source Category # of Mentions

1 @weddady Nasser Weddady Alternative Voices 112 @mathewi Mathew Ingram Mainstream Media 103 @jilliancyork Jillian C. York Alternative Voices 94 @ifikra Sami Ben Gharbia Alternative Voices 75 @natashatynes Natasha Tynes Mainstream Media 76 @nawaat Nawaat de Tunisie Alternative Voices 77 @wjchat “wjchat” Mainstream Media 78 @storify “Storify” Institutional Elites 69 @ethanz Ethan Zuckerman Institutional Elites 510 @monaeltahawy Mona Eltahawy Mainstream Media 511 @jeffjarvis Jeff Jarvis Institutional Elites 312 @lukebozier Luke Bozier Institutional Elites 313 @marocmama Amanda Mouttaki Alternative Voices 314 @tunisianfreedom “Tunisian Freedom” Alternative Voices 315 @amsika Antoine Msika Institutional Elites 216 @basti Bastian Lehmann Other 217 @brian_whit Brian Whitaker Mainstream Media 218 @butwait Shelley Krause Other 219 @charliejane Charlie Jane Anders Alternative Voices 220 @chrisheuer Chris Heuer Institutional Elites 221 @digiphile Alex Howard Institutional Elites 222 @dima_khatib Dima Khatib Alternative Voices 223 @evgenymorozov Evgeny Morozov Institutional Elites 224 @f6x Stephen R. Fox Institutional Elites 225 @globalvoices “Global Voices” Alternative Voices 2


approach. The results of these tests, found in Table 7, suggest that chance cannot be ruled out in thedifferences observed between Alternative Voices and Mainstream Media across both RT and non-RT,even though the original mean and median observations indicate that Carvin preferred atypical sourcesduring the Egypt period. However, the differences between Alternative Voices and Institutional Eliteswere significant for both clusters, reinforcing the observation that nontraditional sources were favoredrelative to elite actors who typically dominate the sourcing patterns of journalists.

Discussion

The routines used by journalists to sort out fact from fiction are rooted in evaluating the credibility ofa source based on assumptions about power, legitimacy, and authoritativeness. The result is a hierarchy


Table 7 Descriptive Statistics for Sourcing Activity, According to Country and Source Type

Period Sourcing ActivitySourceType N Mean Med. Std. Dev. Max. Sum 𝜒2

Groupmean rank

Egyptianrevolution

Retweets(Broadcasting)

MainstreamMedia

62 13.94 3.5 18.48 99 864 16.68 *** 70.72a

InstitutionalElites

32 6.97 3.5 10.84 56 223 48.59

AlternativeVoices

47 29.79 22 32.30 161 1400 86.63a

Non-retweets(Engaging)

MainstreamMedia

52 4.38 3 4.20 18 228 6.91 * 52.49a

InstitutionalElites

26 3.92 3.5 2.62 13 102 53.56

AlternativeVoices

38 6.55 5 5.45 26 249 70.11a

Tunisianrevolution

Retweets(Broadcasting)

MainstreamMedia

26 1.46 1 1.104 6 38 1.26 36.83

InstitutionalElites

24 1.33 1 0.637 3 32 36.96

AlternativeVoices

26 1.73 1 1.282 6 45 41.60

Non-retweets(Engaging)

MainstreamMedia

22 2.36 1 2.517 10 52 .30 41.73

InstitutionalElites

40 1.68 1 1.071 6 67 41.01

AlternativeVoices

21 2.76 1 3.015 11 58 44.17

Notes.1. The “Other” source-type classification has been dropped from this stage of the analysis.2. Sum = number of tweet mentions3. 𝜒2 corresponds to the chi-square statistic on a Kruskal-Wallis test. * = p < .054. Group mean ranks correspond to a Kruskal-Wallis test. Means within each distinct sourcing activity cluster that share no commonsubscript differ at p<.05 by Bonferroni procedure (i.e., p<.0167), according to Mann-Whitney U post-hoc tests. Because neither ofthe Tunisia clusters was significant, pairwise comparisons among the three groups were not considered.

of sources that favors those in positions of power and shapes how issues and events are subsequentlyrepresented in the media. Our study shows that in the cases of both Tunisia and Egypt, mainstream mediaand institutional elites accounted for a substantial percentage of Carvin’s sources. With Egypt, journalistsand their organizations made up the single largest number of sources, just over 30%, indicative of theproclivity of reporters to follow the work of their rivals. But this is only one side of the story, and apotentially misleading one.

While Carvin was in contact with a considerable number of journalists, our analysis indicated thathe favored nonelite sources, particularly in Egypt. Tweets by nonelite sources accounted for just under50% of all the messages in our sample, meaning that alternative voices had a greater influence over thecontent flowing through Carvin’s Twitter stream than journalists or other elite sources. The prioritygiven to alternative voices is particularly strong in Egypt, making up 50% of the 3,291 tweet mentionsin that period, though our data suggests that he also showed a slight propensity for alternative sourcesin Tunisia. Our findings are at odds with the literature that indicates that journalists rely primarily onelite sources and belittle or discredit alternative voices. Research indicates that mainstream media turnto social media content when there is no alternative (Bruno, 2011). This rationale might apply in the


case of Tunisia, a country traditionally underreported in the Western media, but not in the case of Egyptgiven the number of mainstream media organizations that covered the protests. Overall, a proportionalanalysis of Carvin’s tweet mentions suggest that as he transitioned from covering Tunisia to Egypt, henearly doubled his emphasis on alternative voices, gave slightly more attention to mainstream media,and turned away from institutional elites.

The results lend some support to the premise that social media serve as a means to broaden therange of voices in the news. In his coverage, Carvin gave a higher priority to the messages from citi-zens who were expressing their demands for social change, recording and sharing their experiences onTwitter. Such an approach to reporting could be considered as a more representative form of journal-ism, addressing concerns about an overreliance on powerful, institutional elites as sourcing. But at thesame time, it challenges long-held ethical norms of balance, fairness and objectivity in journalism, asthey point to the considerable impact of one group, nonelite sources, in the construction of the news onCarvin’s Twitter feed.

Our findings raise a question of the potential impact of these messages on the content and tone ofother media reporting, given the large number of journalists, editors, and news outlets who monitoredhis feed. The feed developed into a central node in information cascades on the Arab Spring on Twitterthrough Carvin’s practice of reposting content and referencing its source. Our findings raise questionsas to how far this style of reporting on Twitter can reshape the news, by giving voice to the voicelessand challenging institutional elites. How far Carvin’s preference for what may be considered rebel voicescould have influenced the framing and coverage of the events of the Arab Spring was beyond the scopeof our research, however.

Our analysis found that Carvin relied on relatively distinct sets of leading sources depending on thecountry context and the type of sourcing activity (broadcasting vs. engaging). This difference appears tobe related to the nature of Carvin’s work on Twitter: His retweets, arguably a better measure of journal-istic work because they represent information he deemed worthy to disseminate, reveal an ever-greaterdisproportionate preference for alternative voices, as compared to non-retweets. For example, in the caseof Egypt, more than 70% of the most prominent retweeted sources were alternative voices, and the main-stream media voices on the same top 25 list generally were journalists on the ground (e.g., @benncnn).Compare that to the list of the top 25 non-retweeted sources for Egypt, where the prominent representa-tion of media discussion groups (e.g., @pubmedia and @wjchat) and journalists not on the ground (e.g.,@mathewi) suggests that much of Carvin’s non-retweets included him conversing with others about hiswork. Previous research has found that journalists on Twitter commonly engage in this kind of “jobtalking” (Lasorsa et al., 2012). Because Carvin’s use of Twitter spurred an ongoing metadiscussion onTwitter about digital journalism and his role in re-shaping it (e.g., see Silverman, 2011), we can surmisethat his use of @mentions might be more reflective and conversational in nature—rather than a formof newsgathering as such. The retweet function, therefore, is a better gauge of the actor types that influ-enced his reporting. But ultimately, regardless of the sourcing activity being used, our data show that hispreference for alternative voices relative to elites was demonstrably stronger overall.

Carvin’s use of an emerging communications platform diverged from the established paradigm ofjournalistic reporting and its reliance on official sources. In contrast, a study of the coverage of the Egyp-tian uprising in six mainstream newspapers found that journalists preferred to rely on more conventionalsources, rarely citing social media sources (AlMaskati, 2012). Studies into the adoption of new mediaby the mainstream media point to a process of normalization, as journalists transfer established waysof working to digital technologies (Lasorsa et al., 2012; Singer, 2005; Singer et al., 2011). In his analysisof how international news outlets such as the BBC, the Guardian and the New York Times were usingsocial media, Newman (2009, p. 39) concluded that “so far at least, the use of new tools has not led toany fundamental rewrite of the rule book—just a few tweaks round the edges.” Our results suggest that


Carvin adapted journalistic practices in his reporting of the uprising using Twitter, at least when it cameto the selection of sources, rather than transferring existing practices to new tools.

This turn against the paradigm of new media normalization, evident in Carvin’s preference for alter-native voices, can be seen as a revised form of gatekeeping, lending support for Shoemaker and Vos’(2009) proposed model that better accounts for the digital media environment. In this rendering, thereis a more direct and dialogical relationship between the media gatekeeper and his audience, and audi-ences play a key role in enlarging and diversifying the range of source material available to the gatekeeper.Altogether, this contributes to a milieu that Bruns (2005, p. 2) describes as “gatewatching,” signaling ashift from merely publishing newsworthy content to publicizing of relevant information elsewhere, aswell as the evaluation of such external information by means of participation from audience members.Our study suggests that Carvin engaged in gatewatching by pointing his followers to source materialprovided by a diverse set of actors online, contributing to a “real-time verification system” (Silverman,2011, para. 1).

While we did not analyze Carvin’s selection process, the literature indicates that journalists tend tocite sources they are familiar with, as past reliability is taken as an indicator of credibility. Our findingsindicate that Carvin rewarded those involved in documenting events by citing them a significant numberof times in the period analyzed. But the architecture of Twitter may have played a part, too. The retweet-ing functionality means information can be easily amplified and cascade through the network as thereach of a message grows exponentially every time a tweet is rebroadcast. Twitter operates as a sociallyconnected information-sharing network where well-positioned influencers can shape information flows(Kwak et al., 2010). Previous studies have identified how a set of influencers emerged on Twitter duringthe Egyptian uprising. Papacharissi and de Fatima Oliveira (2012) noted the appearance of an organicallycrowdsourced set of opinion leaders of bloggers, activists, and informed citizens on Twitter.

Interestingly, there is some crossover between the alternative sources cited by Carvin and those iden-tified by Papacharissi and de Fatima Oliveira. They include activists such as Mona Seif (@monasosh) andGigi Ibrahim (@gsquare86) who rose to prominence by documenting the protests. The overlap raisesquestions about how and when certain voices rise above others through a process of preferential attach-ment on Twitter. We did not examine the interplay of forces that determine prominence, whether theywere prominent before Carvin cited them or how far the actions of influential actors such as journalistsserved to legitimize these voices.

A further question is how far the selection of sources was shaped by Western interpretations of theevents in Tunisia and Egypt as a struggle for freedom or by Carvin’s specific career track. AlMaskati(2012) found that Guardian International (UK) and International Herald Tribune (US) tended to citemore pro-opposition sources than progovernment ones, suggesting they were following a lead set byWestern officials who regarded the protests as peaceful and democratic. Carvin may similarly have beeninfluenced. Furthermore, the professional background of the NPR social media strategist is far from thepath followed by the traditional foreign correspondent. Before joining NPR in 2006, he worked on issuesof technology, education, and community (Farhi, 2011). Given his background, Carvin may have alreadyhad an inherent predisposition towards alternative voices. In his professional work before NPR, he grewfamiliar with the Global Voices network of bloggers that represents voices usually excluded from themainstream media and developed a large number of contacts (Janssen, 2012).

These questions point to some limitations of this study. Obviously, to engage a more thorough studyof journalistic sourcing and its evolution via social media would require looking beyond the case of AndyCarvin alone—salient though it is—to examine other journalists actively conducting much of their workin these spaces online. Moreover, a networked analysis that more richly captured the spatial dynamicsof information-sharing and user-to-user influence in a social medium like Twitter would reveal nuancesthat cannot be accounted for entirely through traditional research methods such as quantitative and


qualitative analyses of text. To the extent that such data—on a greater number of journalists and sources,across a wider number of critical incidents like the Arab Spring, all represented in a more networkedfashion—can be obtained and analyzed, future research can work to build on this case study of Carvin.

Conclusion

The role of media, especially networked digital technologies, in the wave of protests across the MiddleEast has provoked much debate. Cottle (2011) has noted that media and communication are “inextri-cably infused” (p. 648) with the Arab Spring, while Castells (2011) talked of the birth of “a new systemof mass communication built like a mix between an interactive television, Internet, radio and mobilecommunication systems. The communication of the future has already been used by the revolutions ofthe present” (n.p.). Andy Carvin offers a case study of how a media professional operates within sucha complex and fluid media system. Newman describes such professionals as “multi-media writers andcurators, who can synthesize events in real-time with the best of the backchannel from social media”(2009, p. 10). Carvin’s use of Twitter, while perhaps unique to him in some respects, points to the inno-vative forms of production that emerge in the initial stages of new communication technologies. Theanalysis of his choice of actor types and the frequency of citation suggest there was a new paradigm ofsourcing at play.

Our study contributes to gatekeeping theory by combining multiple levels of analysis—the indi-vidual (Carvin) and social institutional (sources)—and acknowledging the agency and influence of keyindividual gatekeepers (Shoemaker & Vos, 2009, p. 134). Future studies would do well to take this fur-ther with an assessment of individual gatekeepers’ sourcing routines, a la Mr. Gates (1950), in a digitalmedia context. Our findings point to the emergence of a new style of near real-time gatekeeping, wherejournalists cite a potentially broader set of sources through social media. Other research into real-timereporting, in this case at Guardian.co.uk, found a greater use of primary sources than named mediasources on live blogs compared to more established online hard news formats (Thurman & Walters,2013). The open, distributed, and instantaneous nature of social media raises questions for the way jour-nalists work, as they have limited time to process information and are unlikely to be the first to reportthe news. Papacharissi and de Fatima Oliveira posit that the conditions in which news flows on Twitterindicate the “presence of not just one, but several different paradigms of news reporting and journalism”(2012, p. 24).

Twitter is one of a range of social media technologies that privileges contribution, conversation,community and connectivity, compared to the hierarchical structures within established news organi-zations that set the parameters for most news work. Carvin, himself, has described his Twitter networkas “my editors, researchers & fact-checkers. You’re my news room” (Carvin, 2012). He has said his workis “another flavor of journalism,” seeing himself as “another flavor of journalist” (quoted in Farhi, 2011).Our results lend some support to his comments, pointing to the potential of social media platforms toallow for different approaches to news reporting, reshaping journalism practices and influencing whatis defined as journalism. In light of our findings, Carvin emerges as a central node in a networkedmedia environment—one trusted to authenticate, interpret, and contextualize information flows onsocial awareness streams, drawing on a distributed and networked newsroom where knowledge andexpertise are fluid, dynamic, and hybrid.

Acknowledgements

This research was supported by the Social Sciences and Humanities Research Council of Canada and theOffice of the Vice President for Research at the University of Minnesota.


Notes

1 Andy Carvin left NPR on December 27, 2013: http://www.andycarvin.com/?p= 19982 The dataset obtained from Carvin was a 199-megabyte text file that only offered limited adherence to

CSV conventions, thereby preventing us from simply importing it into a statistical analysis package.After studying the file for patterns, we developed a Python script to isolate each individual tweetbased on a set delimiter and create a separate file that included the date of the tweet (from Twitter),the body text of the tweet (from Twitter), the usernames of the sources mentioned in each tweet(text adjacent to each instance of the “@” character), whether a tweet was a retweet or not (based onthe presence of “RT @”), and, if it was a retweet, the username of the source being retweeted (textadjacent to “RT @”). The resulting file was easier to analyze and import into spreadsheet softwareand statistical packages, and provided the researchers with insight not readily apparent in the datasetprovided by Carvin. This approach is described in greater detail in another paper by Lewis, Zamith& Hermida (2013).

3 The Guardian offers a helpful timeline of these events:http://www.guardian.co.uk/world/interactive/2011/mar/22/middle-east-protest-interactive-timeline.

4 Because Carvin tweeted far more over the course of the Egyptian protests than he did throughoutthe Tunisian demonstrations, we opted to use a proportional threshold that would be indicative ofthe prominence of a source relative to the other sources mentioned in the given time period. Ahigher threshold would have resulted in a disproportionately low number of sources in the Egyptiansample, whereas a lower threshold would have resulted in a disproportionately high number ofsources in the Egyptian sample. Thus, the 0.09% and 0.25% thresholds were found to yield the mostcomparable samples.

5 Only 3 sources appeared on all four top 25 lists: @dima_khatib, @monaeltahawy, @weddady. Eightsources appeared in some capacity (either as RT or non-RT) in both the Egypt and Tunisia list sets:@dima_khatib, @jeffjarvis, @jilliancyork, @mathewi, @monaeltahawy, @sultanalqassemi,@weddady, @wjchat.

6 For example, Ben Wedeman (@benncnn) was the second-most retweeted source for Egypt but didnot make the top 25 of non-RT sources for the same period.

7 The distributions were skewed to the right on account of outliers; even with the removal of theseoutliers, along with other transformations, it was determined that the data would not meet therequirements for parametric statistical tests such as one-way ANOVA.

References

AlMaskati, N. A. (2012). Newspaper coverage of the 2011 protests in Egypt, InternationalCommunication Gazette, 74(4): 342–366.

Atton, C., & Wickenden, E. (2005). Sourcing routines and representation in alternative journalism: Acase study approach. Journalism Studies, 6(3), 347–359. doi:10.1080/14616700500132008

Becker, H. S. (1967). Whose side are we on. Social Problems, 14, 239–247.Bennett, W. L. (1988). News: The politics of illusion. New York: Longman.Boczkowski, P. J. (2010). News at work: Imitation in an age of information abundance. Chicago: The

University of Chicago Press.Brown, J. D., Bybee, C. R., Wearden, S. T., & Straughan, D. M. (1987). Invisible power: Newspaper news

sources and the limits of diversity. Journalism & Mass Communication Quarterly, 64(1), 45–54.doi:10.1177/107769908706400106


Bruno, N. (2011). Tweet first, verify later? How real-time information is changing the coverage ofworldwide crisis events. Oxford: Reuters Institute for the Study of Journalism, University of Oxford.Retrieved from http://reutersinstitute.politics.ox.ac.uk/fileadmin/documents/Publications/fellows__papers/2010–2011/TWEET_FIRST_VERIFY_LATER.pdf

Bruns, A. (2005). Gatewatching: Collaborative online news production. New York: Peter Lang.Bruns, A. (2008). Blogs, Wikipedia, Second Life, and beyond: From production to produsage. New York:

Peter Lang.Carlson, M. (2009). Dueling, dancing, or dominating? Journalists and their sources. Sociology

Compass, 3(4), 526–542. doi:10.1111/j.1751-9020.2009.00219.xCarvin, A. (2012, March 26). “I don’t just have Twitter followers. You’re my editors, researchers &

fact-checkers. You’re my news room. And I dedicate this award to you.” Twitter post. Retrievedfrom http://twitter.com/acarvin/status/184424440757624832.

Castells, M. (2011, February 4). I gelsomini tunisini viaggiano in rete. Internazionale. Retrieved fromhttp://www.internazionale.it/i-gelsomini-tunisini-viaggiano-in-rete/

Cottle, S. (2011). Media and the Arab uprisings of 2011: Research notes. Journalism, 12(5), 647–659.doi:10.1177/1464884911410017

Ericson, R. V., Baranek, P. M., & Chan, J. B. L. (1989). Negotiating control: A study of news Sources.Milton Keynes: Open University Press.

Farhi, P. (2011, April 13). NPR’s Andy Carvin, tweeting the Middle East. The Washington Post.Retrieved from http://www.washingtonpost.com/lifestyle/style/npr-andy-carvin-tweeting-the-middle-east/2011/04/06/AFcSdhSD_story.html

Gans, H. J. (1979). Deciding what’s news: A study of CBS Evening News, NBC Nightly News, Newsweek,and TIME. New York: Pantheon Books.

Hall, S., Critcher, C., Jefferson, T., Clarke, J. N., & Roberts, B. (1978). Policing the crisis: Mugging, thestate, and law and order. London: Macmillan.

Hermida, A. (2010). Twittering the news: The emergence of ambient journalism. Journalism Practice,4(3), 297–308. doi:10.1080/17512781003640703

Hermida, A. (2012). Tweets and Truth: Journalism as a discipline of collaborative verification.Journalism Practice, iFirst, March 27, 2012. doi:10.1080/17512786.2012.667269

Janssen, M. (2012, March 26). @acarvin’s example. Retrieved from http://current.org/tech/tech1206carvin.html

Kwak, H., Lee, C., Park, H., & Moon, S. (2010). What is Twitter, a social network or a news media?Proceedings of the 19th international conference on World wide web, WWW ’10 (pp. 591–600).New York: ACM. doi:10.1145/1772690.1772751

Lasorsa, D. L., Lewis, S. C., & Holton, A. E. (2012). Normalizing Twitter: Journalism practice in anemerging communication space. Journalism Studies, 13(1), 19–36.doi:10.1080/1461670X.2011.571825

Lewis, S. C. (2012). The tension between professional control and open participation: Journalism andits boundaries. Information, Communication & Society, 15(6), 836–866. doi:10.1080/1369118X.2012.674150

Lewis, S. C., Kaufhold, K., & Lasorsa, D. L. (2010). Thinking about citizen journalism: Thephilosophical and practical challenges of user-generated content for community newspapers.Journalism Practice, 4(2), 163–179. doi:10.1080/1461670090315691

Lewis, S. C., Zamith, R., & Hermida, A. (2013). Content analysis in an era of big data: A hybridapproach to computational and manual methods. Journal of Broadcasting & Electronic Media,57(1), 34–52. doi:10.1080/08838151.2012.761702


Lotan, G., Graeff, E., Ananny, M., Gaffney, D., Pearce, I., & Boyd, D. (2011). The revolutions weretweeted: Information flows during the 2011 Tunisian and Egyptian revolutions. InternationalJournal of Communication, 5, 1375–1405.

Newman, N. (2009). The rise of social media and its impact on mainstream journalism. Oxford: ReutersInstitute for the Study of Journalism, University of Oxford. Retrieved fromhttp://reutersinstitute.politics.ox.ac.uk/fileadmin/documents/Publications/The_rise_of_social_media_and_its_ impact_on_mainstream_journalism.pdf

Papacharissi, Z., & de Fatima Oliveira, M. (2012). Affective news and networked publics: The rhythmsof news storytelling on #Egypt. Journal of Communication, 62(2), 266–282. doi:10.1111/j.1460-2466.2012.01630.x

Poell, T., & Borra, E. (2011). Twitter, YouTube, and Flickr as platforms of alternative journalism: Thesocial media account of the 2010 Toronto G20 protests. Journalism, Published Online First,December 16, 2011. doi:10.1177/1464884911431533

PRWeb. (2012, March 27). Winners announced after the record breaking 1.6 million tweetnominations for the 4th Annual Shorty Awards Ceremony. Retrieved from http://www.prweb.com/releases/2012/3/prweb9334573.htm

Reich, Z. (2011). Source credibility and journalism. Journalism Practice, 5(1), 51–67. doi:10.1080/17512781003760519

Riffe, D., Lacy, S. R., & Fico, F. G. (2005). Analyzing media messages: Using quantitative content analysisin research (2nd ed.). Mahwah, NJ: Lawrence Erlbaum Associates.

Sakaki, T., Okazaki, M., & Matsuo, Y. (2010). Earthquake shakes Twitter users: Real-time eventdetection by social sensors. Proceedings of the 19th international conference on World wide web,WWW ’10 (pp. 851–860). New York: ACM. doi:10.1145/1772690.1772777

Scott, W. A. (1955). Reliability of content analysis: The case of nominal scale coding. Public OpinionQuarterly, 19(3), 321–325. doi:10.1086/266577

Shoemaker, P. J. (1991). Gatekeeping. Newbury Park: Sage.Shoemaker, P. J., & Reese, S. D. (1996). Mediating the message: Theories of influences on mass media

content. White Plains, NY: Longman.Shoemaker, P. J., & Vos, T. P. (2009). Gatekeeping theory. New York: Routledge.Sigal, L. V. (1973). Reporters and officials: The organization and politics of newsmaking. Lexington,

Mass.: D. C. Heath.Silverman, C. (2011, April 8). Is this the world’s best Twitter account? Columbia Journalism Review.

Retrieved from http://www.cjr.org/behind_the_news/is_this_the_worlds_best_twitter_account.php

Singer, J. B. (2005). The political J-blogger: “Normalizing” a new media form to fit old norms andpractices. Journalism, 6(2), 173–198. doi:10.1177/1464884905051009

Singer, J. B., Hermida, A., Domingo, D., Heinonen, A., Paulussen, S., Quandt, T., … Vujnovic, M.(2011). Participatory journalism: Guarding open gates at online newspapers. New York:Wiley-Blackwell.

Thurman, N., & Myllylahti, M. (2009). Taking the paper out of news: A case study of Taloussanomat,Europe’s first online-only newspaper. Journalism Studies, 10(5), 691–708.doi:10.1080/14616700902812959

Thurman, N., & Walters, A. (2013). Live blogging: digital journalism’s pivotal platform? A case study ofthe production, consumption, and form of Live Blogs at Guardian.co.uk. Digital Journalism, 1(1).

Tuchman, G. (1978). Making news: A study in the construction of reality. New York: Free Press.Twitter. (2012). Twitter Blog: Twitter turns six. March, 21, 2012. Retrieved from http://blog.twitter.

com/2012/03/twitter-turns-six.html


Twitter. (n.d.). About. Retrieved June 10, 2011, from http://twitter.com/about.White, D. M. (1950). The gatekeeper: A case study in the selection of news. Journalism Quarterly, 27,

383–396.

About the Authors

Alfred Hermida ([email protected]) is an Associate Professor in the Graduate School of Journal-ism at the University of British Columbia. A former BBC journalist of 16 years, his research focuses onchanges in journalistic practices, social media and emerging genres of journalism. In 2011, he coau-thored Participatory Journalism: Guarding Open Gates at Online Newspapers (Wiley-Blackwell).Address: UBC Graduate School of Journalism, 6388 Crescent Road, Vancouver, B.C. V6T 1Z2, Canada

Seth C. Lewis ([email protected]) is an Assistant Professor in the School of Journalism & Mass Commu-nication at the University of Minnesota–Twin Cities. His research focuses on the intersection of mediasociology, journalism innovation, and digital technologies. He explores the changing dynamics of jour-nalism’s professional norms, boundaries, and practices. He is a former Fulbright Scholar and a formerjournalist at The Miami Herald.Address: 111 Murphy Hall, 206 Church Street SE, Minneapolis, MN 55455, USA

Rodrigo Zamith ([email protected]) is a doctoral student in the School of Journalism and Mass Com-munication at the University of Minnesota–Twin Cities. His research interests include computationalresearch methods in content analysis, journalism and new media technologies, and the role of newsmedia in framing issues. He has previously worked at the Minneapolis Star Tribune and the St. Paul Pio-neer Press.Address: 111 Murphy Hall, 206 Church Street SE, Minneapolis, MN 55455, USA


Sourcing the Arab Spring: A Case Study of Andy Carvin’s...

Documents

Transcript of Sourcing the Arab Spring: A Case Study of Andy Carvin’s...